Gut bacteria prime and reinforce mature leaf oviposition preference in beetles via plant mediated feedback Authors: Chong Li, Ziyue Sun, Yuxin Zhang, Meiqi Ma, Letian Xu Published: 2026 ABSTRACT AbstractThe gut microbiota can profoundly influence host behavior, yet underlying mechanisms for microbiota-driven behavioral adaptation remain elusive. We previously observed that female Plagiodera versicolora leaf beetles preferentially oviposit on mature leaves despite feeding on new leaves. Here, we demonstrate that this oviposition preference is gut microbiota-dependent, as it is abolished by microbiota depletion and restored by microbiota transplantation. Gut Pseudomonas putida emerges as the key driver, with its abundance positively correlating with mature-leaf preference. Further investigation reveals that beetle feeding induces elevated indole-3-acetic acid (IAA) levels, promoting bacterial translocation from feeding sites to new leaves. This colonization facilitates bacterial dissemination while reducing F1 fitness during larval feeding. Comparative transcriptomics implicates the takeout gene (TO1) in regulating oviposition preference, and TO1 knockdown abolishes this behavior. Moreover, P. putida supplementation restores TO1 expression and mature-leaf preference in germ-free females. These findings reveal an insect–microbe–plant feedback loop driving microbiota-associated behavioral adaptation.Subject terms: Microbial ecology, Behavioural ecology, Microbiome, Chemical ecology Introduction Insect behavior is a complex process regulated by the interplay of multiple factors, including visual and olfactory signals, as well as intricate internal neural pathways1–4. Ultimately, these processes manifest as a series of life-history decisions—such as foraging strategies, mate selection, and habitat colonization—that aim to maximize the individual’s or their offspring’s fitness. Among these decisions, oviposition site selection is particularly critical: once eggs are laid, they cannot relocate, leaving embryos and newly hatched larvae vulnerable to biotic threats (e.g., parasitoids, predators, and host plant defenses) and abiotic stresses (such as temperature, humidity)5,6. Moreover, the choice of oviposition substrate determines the quality and quantity of resources available to the larvae, directly affecting larval survival and development5,7,8. Understanding the factors that guide oviposition decisions and deciphering the mechanisms are therefore of great importance not only for advancing our understanding of insect-plant interaction and life-history strategies but also for the development of strategies in both beneficial insect conservation and pest management, thus attracting substantial research interest. Recent studies have revealed that microorganisms—ranging from beneficial symbionts to opportunistic pathogens—can influence nearly every aspect of insect reproductive behavior. These microbes modulate behaviors such as mate finding, courtship, and oviposition-site choice by producing or altering chemical cues in the host insect’s environment9–11. For example, in Drosophila, environmental microbes release volatile compounds that alter female egg-laying rates and substrate preferences12; in mosquitoes, the presence of specific bacteria signals the suitability of oviposition water13. Beyond signaling the potential risk of infection or parasitic threat at oviposition sites, insects can gauge food availability or detect detrimental microorganisms via microbial cues; for instance, surface bacteria on eggs of Bactrocera dorsalis stimulate plants to emit β-caryophyllene, which deters additional females from ovipositing at the same site14. Compared to environmental microbiota, gut-resident microbes maintain an intimate and often stable association with their hosts9,15. These resident communities encompass a diverse spectrum of interactions, ranging from beneficial mutualists to antagonistic or opportunistic pathogens16,17. A growing body of research demonstrates that specific members of these microbial communities can influence various insect behaviors, often mediated through pathways like the gut-brain axis18–21. However, whether and how gut bacteria directly influence oviposition site selection in vivo remains largely unexplored. An additional significant puzzle for researchers is how key gut symbionts overcome barriers to horizontal and vertical transmission, particularly in non-social insects, to maintain their influence across hosts’ life stages and generations. The willow leaf beetle, Plagiodera versicolora, is a significant pest of willow and poplar trees22. To better exploit the limited leaf resources, the beetle has evolved an ecological strategy of dietary specialization. In this context, the term encompasses both host-plant specificity—which reduces interspecific competition—and ontogenetic niche partitioning, characterized by the distinct leaf tissue preferences of adults and larvae, which minimizes intraspecific competition. Specifically, although adult beetles preferentially feed on new leaves23,24, females predominantly oviposit on mature leaves—a strategy that spatially separates egg deposition sites from adult feeding sites25,26. More importantly, the phenomenon aligns with the larvae’s specialized adaptation for feeding on mature leaves16,24,26. While this adult preference for mature-leaf oviposition is a crucial prerequisite for niche partitioning and dietary specialization, thereby enhancing fitness, the underlying mechanisms driving this specific oviposition choice remain unclear. In this study, we demonstrate that gut bacteria play a pivotal role in shaping the adult preference for ovipositing on mature leaves. Treatment with antibiotics markedly altered the wild population females’ tendency to lay eggs on mature leaves; germ-free (GF) females shifted their oviposition preference toward new leaves, whereas reintroduction of gut microbiota into GF females fully restored their mature-leaf preference. We then identified Pseudomonas putida as a key gut bacterium positively correlated with mature-leaf oviposition preference, which can actively translocate to new leaves from feeding sites in response to plant indole-3-acetic acid (IAA) gradients. Given adult beetles’ preference for feeding on new leaves, this bacterial accumulation on those leaves not only promotes their dispersal but also reinforces the female propensity to oviposit on mature leaves, establishing an insect-microbe-plant feedback loop. At last, we revealed P. putida influences this preference by modulating the expression of the takeout gene in P. versicolora. While the precise mechanism by which P. putida regulates takeout requires further investigation, our findings unveil a previously unappreciated mode and mechanism of plant-insect-gut microbiota interaction that shapes host behavior. Results Gut microbiota mediates oviposition preference independent of feeding capacity and reproductive fitness Sampling of Plagiodera versicolora populations was conducted in Qinhuangdao (Hebei, China), Wuhan (Hubei, China), and Guiyang (Guizhou, China), alongside field experiments at Shahu Lake in Wuhan (Fig. 1A). Across all the populations, we consistently observed that the beetles preferred to oviposit on mature leaves; however, this preference for mature leaves was significantly altered when gut microbiota were depleted using antibiotics (Figs. 1B, C and S1A, B). Besides, no significant differences in leaf consumption over 24 h were observed between CONV and ABX females across the three geographical populations (Fig. S1E). To mitigate potential confounding factors associated with field conditions or antibiotic treatments, further validation was performed using laboratory-established germ-free females. Specifically, we established three experimental groups—conventionally reared (CONV) females, germ-free (GF) females, and microbiota-reintroduced (GF + GM) females—to rigorously assess the role of gut bacteria in oviposition choice (Fig. 1D). We also verified the axenic status of the GF beetles through LB medium culture and qPCR techniques, and the results indicated no gut bacteria were detected in the GF adults (Fig. S1C, D). Behavioral assays showed that mated CONV females significantly preferred mature leaves over new leaves for egg-laying (Fig. 1E, PI = 0.37 ± 0.08). In contrast, GF females exhibited a significant shift in their oviposition preference towards new leaves (PI = −0.49 ± 0.09). Re-introducing of gut microbiota (GF + GM) restored their preference for mature leaves, similar to the CONV group (Fig. 1E, PI = 0.30 ± 0.07). Regarding other reproductive parameters, there was no significant difference in the size of individual egg clutches deposited by females between mature and new leaves across the three treatment groups (Fig. 1F). Furthermore, the hatching success of eggs was similar between the two leaf types across all treatment groups, showing no significant differences (Fig. 1G). Gut microbiota modulate adult oviposition preference on mature versus new leaves. A Colored text labels indicate the specific sampling sites: Qinhuangdao (QHD) in Hebei Province (Green), Wuhan (WH) in Hubei Province (Blue), and Guiyang (GY) in Guizhou Province (Orange). The scale bar represents 500 km. B Schematic illustration of the field experimental setup. Randomly selected willow branches, featuring both mature and new leaves (as shown in the insets), were enclosed in transparent mesh cages to prevent female escape. Female beetles were then introduced, and their oviposition activity was recorded after 24 h. C Oviposition preference of adults from different populations. Bar charts display the frequency of female oviposition on new versus mature leaves for conventionally reared (CONV) and antibiotic-treated (ABX, gut microbiota-depleted) individuals (n = 30). D Schematic diagram illustrating the experimental procedure. Three groups of adult beetles were established: conventionally reared females (CONV), germ-free females (GF), and germ-free females re-inoculated with gut microbiota (GF + GM). These groups were then subjected to oviposition preference tests. E Oviposition preference index of P. versicolora adults for the three groups when presented with both new and mature leaves (n = 20, left panel). And mature leaf (ML) oviposition rate (n = 20, right panel) F Egg cluster size (number of eggs per clutch) laid on mature and new leaves by each treatment group (n = 30). G Hatching rate (%) of eggs deposited on mature and new leaves (n = 30). Data are presented as mean ± SEM (E right panel, and G). ***p < 0.001 using the two-sided chi-square test for (C) (comparing new versus mature leaf categories within each group), two-sided one-sample t-test for (E) (comparing the PI of each group against a theoretical PI = 0), and two-sided independent samples t-test for (F and G). ns indicates not significant. One-way ANOVA with Tukey’s post hoc multiple comparisons for (E) (right panel) and different letters indicate statistically significant differences at p < 0.05. All p-values are listed in Supplementary Dataset 3. Diet-independent stability of the gut microbiota and its correlation with oviposition preference Because diet is a primary driver of gut microbial composition, we reared adult P. versicolora females on either new (NLF) or mature leaves (MLF), then analyzed their gut communities by 16S rRNA sequencing and measured oviposition behavior. Simultaneously, we surveyed the epiphytic microbiota of healthy willow leaves (both new [NL] and mature [ML]) from three distinct sites in Wuhan (Jiangxia, Huangpi, and Wuchang) to provide an ecological baseline. Oviposition assays showed that, regardless of whether they were fed new or mature leaves, mated females still strongly preferred to lay eggs on mature leaves, and there were no significant differences between new or mature leaves in egg clutch size or hatching rate (Fig. 2A–C; PI for NLF = 0.31 ± 0.11; PI for MLF = 0.33 ± 0.10). Microbial alpha-diversity analysis indicated that gut communities of NLF and MLF adults possessed comparable richness and evenness, as reflected by the Chao1, Simpson, and Shannon indices (Fig. 2D–F). And leaf microbiota diversity was generally similar between new and mature leaves across three sites (Fig. 2D–F). Beta-diversity analysis showed that gut communities from NLF and MLF adults clustered closely together, indicating high stability of the adult gut microbiota against dietary shifts (Fig. 2G). In contrast, leaf-associated communities clustered primarily by geographic site rather than leaf age, and remained distinct from the gut microbiota (Fig. 2G). Effects of adult feeding diet on oviposition preference and gut microbiota composition, and correlation between mature leaf oviposition preference and the abundance of gut bacteria. A Left panel: Oviposition preference index of P. versicolora adults that had been fed new leaves (NLF) or mature leaves (MLF). Each bubble represents a specific preference index value, with its size indicating the number of females exhibiting that preference (number inside the bubble) (n = 20). Right panel: mature leaf oviposition rate (n = 20). B Number of eggs per cluster laid on mature and new leaves by females fed new leaves (NLF) or mature leaves (MLF) (n = 30). C Hatching rate (%) of eggs deposited on mature and new leaves by the two diet groups (n = 30). D–F α-diversity indices of gut microbiota in MLF and NLF adults, and leaf microbiota from three locations (JX Jiangxia, WC Wuchang, HP Huangpi): D Chao1 richness estimator, E Shannon index, and F Simpson index. For the three panels, n = 7 for NLF, and n = 5 for all other groups. In the box plots, the center line indicates the median, the box limits represent the 25th and 75th percentiles, and the whiskers extend to the minimum and maximum values. G Principal Coordinates Analysis (PCoA) of gut and leaf microbiota profiles based on Bray-Curtis distance (PC1 = 44.14%, PC2 = 12.97%) (n = 5 or 7). H Relative abundance of gut and leaf microbiota at the genus level. I UpSet plot showing the shared and unique amplicon sequence variants (ASVs) across adult gut (MLF, NLF) and leaf (pooled M_leaf and N_leaf from three locations) communities. J–M Pearson’s correlations between the mature leaf (ML) oviposition rate and the relative abundance (top row, n = 12) or absolute abundance (bottom row, n = 18) of four dominant bacterial genera: J Pseudomonas, K Enterobacter, L Pantoea, and M Ralstonia. Solid lines indicate linear regression fits, and shaded area represent 95% confidence intervals. Reported r² and p-values are from Pearson’s correlation. Data are presented as mean ± SEM (A, right panel, and C). Statistical significance was assessed using two-sided one-sample t-test for (A) (left panel), two-sided independent samples t-test for (A) (right panel), and B–F; PERMANOVA for (G). *p < 0.05 or **p < 0.01; ns not significant. All p-values are listed in Supplementary Dataset 3. Taxonomic composition analysis at the genus level showed that the gut microbiota was dominated by Enterobacteriaceae and Pseudomonas, a profile significantly different from the more diverse epiphytic leaf microbiota (Fig. 2H). To identify key candidates involved in behavioral modulation, we utilized an UpSet analysis to track shared and unique taxa. While a majority of ASVs were specific to leaves, we identified a core set of ten candidate genera, including 8 genera shared across all groups (ML, NL, MLF, and NLF) and two genera exclusively found in the gut (Fig. 2I). Correlation analysis between the abundance of these 10 candidates and the mature-leaf (ML) oviposition rate revealed that only Pseudomonas exhibited a significant positive correlation with the behavioral preference (Fig. 2J). In contrast, the other nine genera examined, such as Enterobacter and Pantoea, were not significantly correlated with this preference (Figs. 2K–M and S2). Quantitative PCR of Pseudomonas 16S rRNA genes validated the sequencing results, confirming that elevated Pseudomonas abundance is associated with the mature-leaf oviposition preference (Fig. 2J). IAA-driven translocation of P. putida impairs F1 larval performance when feeding Screening of Pseudomonas isolates from the beetle gut identified six distinct species, among which P. putida emerged as the most abundant (45.7%, Fig. 3A) and was therefore selected for further functional characterization. Herbivory can induce IAA-mediated compensatory growth and chemical defense responses27–29. Given the reported chemotaxis of Pseudomonas toward IAA30, we quantified the spatiotemporal dynamics of IAA, jasmonic acid (JA), and chlorogenic acid (CGA), in parallel with P. putida abundance over a 48-h period in upper leaves following P. versicolora feeding or mechanical wounding (Fig. 3B–F). Heatmap analysis demonstrated that feeding by P. putida-supplemented CONV beetles (CONV + P. p) significantly elevated IAA levels in leaves immediately above the feeding site at 6, 12, and 48 h, while the CONV group exhibited significant increases at 12 and 48 h; notably, a significant increase in IAA was also observed in the ABX group at the 12-h mark (Fig. 3C). This IAA accumulation was tightly synchronized with a significant increase in bacterial fold change (Fig. 3F). In contrast, although mechanical wounding with topical P. putida application (MW + P. p), CONV, and CONV + P. p treatments led to a slight elevation in JA levels, these changes remained statistically insignificant (Fig. 3D). Furthermore, while the defensive metabolite CGA exhibited significant induction starting as early as 6 h in the CONV group and from 12 h in the MW + P. p and CONV + P. p groups, almost no significant changes were observed in the ABX treatment (Fig. 3E). Collectively, these results indicate that the enrichment of P. putida is specifically coupled with plant-mediated IAA induction rather than a generalized mechanical wounding response. Spatiotemporal dynamics of plant chemicals and Pseudomonas putida recruitment, and pathways of horizontal and vertical transmission. A Composition of Pseudomonas species isolated from the adult beetle gut, with P. putida identified as the predominant isolate. B Schematic representation illustrating the different leaf positions sampled for analyses relative to the beetle-fed leaf (BFL). UL upper leaves, BFL beetle fed leaf. C–F Heatmaps showing the spatiotemporal concentrations of indole-3-acetic acid (IAA) (C), jasmonic acid (JA) (D), and chlorogenic acid (CGA) (E), along with relative P. putida fold change (log2) (F) across five treatment groups (CONV, ABX, MW, MW + P. p, and CONV + P. p) over a 48-h period (n = 4–5). G, H Chemotactic response of P. putida (G) and E. aerogenes (H) toward a range of IAA concentrations (n = 5). I, K Plant-mediated horizontal transmission of P. putida. I Schematic of the experimental procedure: donor beetles (B1) infested plants, and naive recipient beetles (B2) were fed on the translocated upper leaves (L2). K Absolute abundance of eGFP copy numbers in L2 leaves and B2 guts (n = 4). J–L Transgenerational vertical transmission of P. putida. J Schematic illustrating the symbiont tracking across life cycle stages. L Absolute quantification of eGFP copy numbers in eggs, larval instars (1st–3rd), pupae, and adult tissues/frass (n = 4). Data are presented as mean ± SEM (G, H, K, L). Statistical significance was determined using one-way ANOVA with Tukey’s post hoc test for (G, H), and two-sided independent t-test for (C–F and K). *p < 0.05, **p < 0.01, or ***p < 0.001. Different letters indicate statistically significant differences at p < 0.05. ns indicates not significant. All p-values are listed in Supplementary Dataset 3. We further conducted in vitro chemotaxis assays using a range of IAA, MeJA, and CGA concentrations using the representative strain. P. putida exhibited no chemotactic response toward MeJA or CGA (Fig. S3). In contrast, P. putida exhibited positive, concentration-dependent chemotaxis towards IAA across a range of 1–300 μM. At a higher concentration (1000 μM), the chemotactic response decreased but remained significantly higher than the control without IAA (Fig. 3G). Another gut commensal, E. aerogenes, showed no significant chemotaxis towards IAA at any tested concentration (Fig. 3H). This phenomenon was further corroborated in field willow trees. Specifically, in new leaves of willow trees fed by adult beetles at three field sites (Guiyang, Wuhan, and Qinhuangdao), IAA content was higher than in new leaves of unfed plants (Fig. S4). Moreover, the abundance of total Pseudomonas and P. putida was positively correlated with IAA content (Fig. S4). To investigate whether upward-translocated P. putida can be transmitted to other feeding beetles, we first colonized the gut of female adults with P. putida harboring an eGFP plasmid (P. putida-eGFP). These females were then kept on willow leaves for feeding. Subsequently, upper sections of these fed leaves were offered to naive female beetles (Fig. 3I). We quantified eGFP copies in the upper leaves and in the gut of the recipient females (Fig. 3K). Results showed that eGFP copies were present in both the upper leaves from P. putida-eGFP damaged plants and in the gut of the females that consumed these sections (Fig. 3K). Fluorescence microscopy showed the physical presence and vascular translocation of eGFP-tagged P. putida within the xylem and internal tissues of the upper leaves and adjacent stem internodes (Fig. S5). Furthermore, we tracked the transmission of the eGFP-tagged P. putida across the life cycle of the F1 generation (Fig. 3J). Results showed that P. putida persists from eggs through all F1 stages, with bacterial loads increasing obviously during larval development and remaining high in the adult gut and frass (Fig. 3L), confirming the robust transgenerational stability. To further evaluate the ecological significance of this P. putida migration, we assessed larval performance by feeding them leaves collected from the upper and lower leaves of plants previously subjected to the different feeding treatments (CONV female-fed, ABX female-fed, and unfed controls, Fig. 4A). Compared to larvae fed leaves from ABX female-fed or unfed control plants, larvae fed upper leaves from CONV female-fed plants exhibited significantly reduced performance, including lower third instar larval weight (7.74 ± 0.22 mg), lower pupal weight (7.80 ± 0.19 mg), and slower development time (1st instar: 45.24 ± 0.50 h; 2nd instar: 62.27 ± 0.89 h; 3rd instar: 72.08 ± 1.26 h) (Fig. 4B–D). In contrast, the performance of larvae fed lower leaves was similar among the three treatment groups, including third instar larval weight (control: 8.44 ± 0.28 mg), pupal weight (control: 8.82 ± 0.20 mg), and larval development time (control: 1st instar: 43.14 ± 0.30 h; 2nd instar: 59.66 ± 0.41 h; 3rd instar: 68.89 ± 0.60 h) (Figs. 4B, C, and E). Consistently, qPCR analysis revealed that P. putida specifically accumulated in the upper leaves of plants fed on by conventional beetles (Fig. 4F), and a significant increase in P. putida abundance within the larvae that consumed these upper leaves (Fig. 4G). Furthermore, we found that larvae showed no significant feeding preference between leaves coated with P. putida and control leaves (Fig. 4H, PI = −0.06 ± 0.10). To distinguish between direct bacterial effects and plant-mediated responses, we performed direct infection assays via leaf-surface application. The results confirmed that P. putida significantly reduced larval survival but had no impact on adult survival, fecundity, or egg hatching rates (Figs. 4I and S6). Impact of Pseudomonas putida accumulation on larval performance and stage-specific susceptibility. A Schematic diagram illustrating the experimental setup. Larvae were reared on detached upper (UL) or lower (LL) leaves from plants that were either unfed control or previously fed upon by CONV or ABX P. versicolora adults (Beetle Fed Leaf, BFL). B, C body weight of 3rd instar larvae and pupa after feeding on the UL or LL leaves from the indicated plant treatments (unfed control, CONV-fed, ABX-fed) (n = 30). D, E developmental time of larvae fed on UL or LL from the indicated plant treatments (unfed control, CONV-fed, ABX-fed) (n = 30); The color gradient from light to dark within each bar represents the developmental duration of the 1st, 2nd, and 3rd larval instars, respectively. F, G Relative expression of P. putida 16S rRNA in UL/LL tissues and larvae. Data are shown for leaf tissues (F, n = 6, with one leaf per sample) and larvae fed on those respective leaves (G, n = 4, with one larva per sample), quantified by qPCR. H Feeding preference index (PI) of P. versicolora larvae in a choice test between leaves coated with P. putida suspension (104 CFU/cm²) and control leaves coated with ddH₂O, with the number inside indicating the count of larvae within that PI range (n = 15). I Survival curves showing the probability of survival for larvae and adults fed on detached leaves coated with P. putida (1 × 104 CFU/cm²) compared to CONV controls (n = 30). Data are presented as mean ± SEM (B–G). Statistical significance was determined using one-way ANOVA with Tukey’s post hoc test for (B–G), two-sided one-sample t-test for (H), and Log-rank (Mantel–Cox) test for survival curves in panel I. Different letters indicate statistically significant differences at p < 0.05. **indicates p < 0.01; ns indicates not significant. All p-values are listed in Supplementary Dataset 3. P. putida-driven regulation of a takeout gene modulates oviposition preference To pinpoint how gut microbiota influences oviposition preference in P. versicolora, we compared the pre-oviposition head transcriptomes of GF and CONV females. Principal Component Analysis (PCA) revealed a clear separation between the two groups (Fig. 5A). Further analysis revealed 61 differentially expressed genes (DEGs), which were predominantly associated with structural proteins (e.g., cuticle and chitin-binding proteins), immune response factors, and oxidoreductases (Supplementary Data 2). Interestingly, two DEGs involved in neural signaling were identified. One takeout homolog (TO1, Pver133270)—linked to neural signaling—was significantly downregulated in GF females relative to CONV controls (Fig. 5B). qPCR validation showed that TO1 expression levels positively correlate with the degree of mature-leaf oviposition preference (Fig. 5C), whereas a neuropeptide-like (NPL, Pver004360) gene exhibited an inverse relationship (Fig. 5D). Spatiotemporal profiling showed that TO1 expression peaks specifically in the head during the pre-oviposition phase (Fig. 5E). In contrast, NPL exhibited a peak in the body tissues during the same period (Fig. 5F). Structural and phylogenetic analysis confirmed that TO1 possesses the hallmark features of the “takeout” family, including a signal peptide, a juvenile hormone-binding protein (JHBP) domain, and two conserved cysteine residues (Fig. 5I). Consistent with this, RNAi knockdown of TO1 disrupted the normal preference for mature leaves (Fig. 5G, PI for dsTO1 = −0.34 ± 0.15, PI for dsGFP = 0.27 ± 0.09). TO1 expression in the dsTO1-treated group was significantly suppressed, mirroring the levels observed in GF females (Fig. 5H). Transcriptomic and functional analyses of two genes related to neural function in conventional (CONV) and germ-free (GF) female Plagiodera versicolora. A Principal Component Analysis (PCA) of head transcriptomes from conventional (CONV) and germ-free (GF) females, calculated using the top 2000 most variable genes (n = 3). B Volcano plot of head transcriptomes comparing CONV and GF females. The x-axis represents log2(fold change) and the y-axis log10 (p-value); yellow and blue dots denote genes significantly up or downregulated in CONV versus GF (|log2FC| ≥ 1, FDR < 0.05, TPM > 10), respectively, and grey dots non-significant genes. Two candidate DEGs (TO1 and NPL) are highlighted. C, D Correlation between relative expression levels of TO1 (C) and NPL (D) and the proportion of eggs laid on mature leaves (n = 9–10 per group, total n = 19 for the correlation). Solid lines indicate linear regression fits, and shaded area represent 95% confidence intervals. E, F Spatiotemporal qRT-PCR expression profiles of TO1 (E) and NPL (F) in head and body tissues at three physiological stages: emergence, pre-oviposition, and post-oviposition (n = 3 or 4; precise values indicated by individual data points). Each sample represents a pool of 10 heads or 3 bodies. G Left panel: effect of dsRNA-mediated TO1 knockdown on oviposition preference index of females when presented with both new and mature leaves. CONV females were injected with dsTO1 (n = 10) or control dsGFP (n = 20). Right panel: Mature Leaf (ML) oviposition rate for the two treatment groups. H Head expression levels of TO1 in CONV, GF, and dsTO1-treated females (n = 4, each sample represents a pool of 10 heads). I Conserved domain architecture and phylogenetic analysis of TO1 orthologs from eight representative insects. Left: maximum likelihood phylogenetic tree with bootstrap support values at nodes. Right: protein schematics to scale showing signal peptide (red), juvenile hormone binding protein domain (JHBP) domain (blue), and conserved cysteine residues (yellow triangles). Data are mean ± SEM (E, G right panel, H, F). Statistical significance was assessed using one-way ANOVA with Tukey’s post hoc test (E, F, H), two-sided one-sample t-test (G, left, comparing the PI of each group against a theoretical PI = 0), and two-sided independent samples t-test (G, right). *p < 0.05 or **p < 0.01; different lowercase letters denote significant differences at p < 0.05, ns indicates not significant. All p-values are listed in Supplementary Dataset 3. To determine whether P. putida alone is sufficient to modulate host behavior, we reintroduced either P. putida or a control gut bacterium, Enterobacter aerogenes, into GF females (Figs. 6A and S9). Reintroduction of P. putida successfully reversed the preference shift observed in GF females, establishing a significant preference for mature leaves (Fig. 6B, PI for GF = −0.39 ± 0.09, PI for GF + P.p = 0.26 ± 0.10). In contrast, E. aerogenes colonization failed to rescue this behavioral phenotype (Fig. 6B, PI = −0.15 ± 0.12). Corroborating these behavioral shifts, qPCR analysis showed that P. putida, but not E. aerogenes, significantly upregulated TO1 expression compared to GF controls (Fig. 6C). We next examined the functional requirement of TO1 in P. putida-mediated behavioral restoration by performing RNAi-mediated knockdown prior to bacterial reintroduction. Despite the administration of dsTO1, reintroduction of P. putida significantly increased TO1 transcript levels relative to the dsTO1 control (Fig. 6E), which was sufficient to restore the mature-leaf preference (Fig. 6D, PI for GF+dsTO1 = −0.64 ± 0.13, PI for GF + dsTO1 + P.p = 0.45 ± 0.10). Conversely, E. aerogenes had no significant effect on either TO1 expression or oviposition choice under knockdown conditions (Fig. 6D, E, PI = −0.46 ± 0.07). Finally, supplementing CONV females with P. putida further increased P. putida titers in the gut, which correlated with significantly elevated TO1 expression and an intensified oviposition preference for mature leaves (Fig. 6A, F–H). Reintroduction of Pseudomonas putida modulates TO1 expression and contributes to the reinforcement of mature leaf oviposition preference. A Schematic representation of the experimental procedure. Germ-free (GF) P. versicolora adults were treated with either H₂O (control) or dsTO1, and then re-associated with P. putida (P. p) or E. aerogenes (E. a) or kept germ-free. Conventional adults were supplemented with H2O or P. putida. Subsequent adult oviposition preference tests were conducted using detached mature and new leaves. B Oviposition preference index (left panel) and mature leaf (ML) oviposition rate (right panel) of untreated GF P. versicolora adults, GF adults reintroduced with P. putida (GF + P. p), or with E. aerogenes (GF + E. a). The size of each circle is proportional to the number of egg clusters laid within that preference index range, with the exact number indicated inside the circle (n = 10 or 20). D Oviposition preference index (left panel) and mature leaf (ML) oviposition rate (right panel) of dsTO1-treated germ-free adults (GF + dsTO1), GF + dsTO1 reintroduced with P. putida (GF + dsTO1 + P. p) or with E. aerogenes (GF + dsTO1 + E. a) (n = 10). C, E Gene expression level in different treatments (C, n = 10, 10, 6; E, n = 8, 6, 4 from left to right. Each replicate represents a pool of 6 adult heads). F Oviposition choice frequency on new versus mature leaves for CONV females treated with H2O or supplemental P. putida. Numbers inside bars indicate the number of females that chose each leaf type (n = 60). G, H Relative abundance of (G) P. putida (16S rRNA) (n = 4, 3 guts per replicate) and H TO1 expression in CONV + H2O and CONV + P. p females (n = 4, 8 heads per replicate). Data are presented as mean ± SEM (right panels of B and D, and C, E, G, H). * p < 0.05, **p < 0.01, or ***p < 0.001 using two-sided one-sample t-test for B (left panel) and D (left panel) (comparing the PI of each group against a theoretical PI = 0); and one-way ANOVA with Tukey’s post hoc test for B (right panel) and D (right panel), (C and E). Two-sided Chi-square test for oviposition choice frequency in (F) (comparing new versus mature leaf categories within each group), and two-sided independent samples t-test for (G and H) (comparing CONV + H2O versus CONV + P. p females). All p-values are listed in Supplementary Dataset 3. Discussion Feeding and oviposition are fundamental life-history traits in insects. While feeding directly impacts individual fitness, oviposition not only contributes to reproductive success but also ensures the transmission of genes to the next generation. These processes are often tightly linked; for example, dietary quality can influence reproductive output28,31. In P. versicolora, this linkage is reflected in dietary specialization and ontogenetic niche partitioning: adults preferentially feed on new leaves but lay eggs on mature leaves. This seemingly counterintuitive oviposition preference can be advantageous, spatially separating larvae from the adult feeding niche and potentially maximizing group fitness26. Prior studies suggested that the preference for feeding on new leaves is primarily nutritionally driven, as consuming new leaves leads to higher egg production24,28,32. Building upon this, our current study delves into the underlying mechanism driving the oviposition preference for mature leaves. While acknowledging the potential involvement of other sensory cues (olfactory, visual)33, the oviposition preference was markedly attenuated after the elimination of gut microbiota in both field and laboratory populations—coupled with the complete restoration of mature-leaf preference upon microbial reintroduction (Fig. 1C and E), strongly indicates that gut bacteria are a crucial driving factor in shaping this behavior. Notably, despite their essential role in oviposition preference, gut microbiota had no detectable effects on fecundity parameters or feeding behavior (Figs. 1F, G and S1E), suggesting that their influence is specific to behavioral decision-making rather than mediated through general physiological condition. Gut microbes can influence insect behavior through various means, including the production of volatile chemicals or direct modulation of host brain transcriptional and neurol processes through gut-brain axis14,34–36. The gut-brain axis, where gut microbes communicate with the host nervous system, is increasingly recognized as a significant regulator of insect behavior19,35,36. Having established the requirement of gut bacteria for oviposition preference, we focused on Pseudomonas putida—the predominant species in the female gut—whose abundance correlates positively with the mature-leaf preference (Figs. 2J and 3A). A fundamental question arising from these findings is how a non-social insect like P. versicolora maintains its association with the gut bacterium and facilitates its transmission? Actually, herbivorous insects may benefit from their host plants, which could serve as a potential vector. For instance, silkworm larvae are known to acquire beneficial bacteria antagonistic to fungal pathogens directly from plant leaves37,38. Theoretically, leveraging the plant environment to enrich and transmit these bacteria may represent a more effective strategy than simple passive acquisition from leaf surfaces, particularly if beetles preferentially feed on plant tissues (e.g., compensatory growth leaves) where these bacteria are enriched25. Herbivory in willow has been shown to induce compensatory growth27,28, often accompanied by the activation of plant defense responses29. Integrated spatiotemporal analysis of plant hormones and defensive metabolites reveals that P. putida enrichment is specifically coupled with localized IAA accumulation rather than a generalized mechanical wounding response (Fig. 3). In vitro assays further demonstrate a chemotactic response of P. putida toward IAA (Figs. 3 and S3), while fluorescence microscopy suggests that this enrichment may involve transport through the plant vascular system (Fig. S5). Although the potential influence of other feeding-induced phytochemicals cannot be entirely excluded, field surveys confirmed a significant positive correlation between IAA levels and Pseudomonas abundance in beetle-damaged leaves (Fig. S4). This plant-mediated migration mechanism, also corroborated by studies such as Wang et al., who found enrichment of Pseudomonas in willow leaves after leaf beetle feeding, though the authors did not distinguish mature and new leaves39. This plant-mediated migration facilitates horizontal transmission to adult beetles preferentially feeding on new leaves (Fig. 3I and K). Furthermore, our eGFP-tagged P. putida assays (Fig. 3J and L) demonstrate that once colonized in the adult gut, P. putida persists in the F1 generation even without additional environmental supplementation, indicating a potential route for vertical transmission. Collectively, this dual transmission strategy maintains the association between P. putida and female beetles, thereby reinforcing the specialized mature-leaf oviposition preference (Fig. 6F–H). On the other hand, this plant-mediated dispersal of P. putida appears to create an ecological deterrent mechanism, which influences female oviposition behavior by negatively impacting the F1 generation fitness. Our previous work identified P. putida as a conditional pathogen that can synergistically enhance the efficacy of RNAi against P. versicolora larvae40. In the current study, we observed a clear fitness cost associated with bacterial enrichment: larvae reared on upper leaves from CONV-infested plants—where P. putida populations are high—exhibited significantly stunted growth and development compared to those on ABX or unfed treatments (Fig. 4). This observed fitness cost is consistent with the hypothesis that the female’s preference for ovipositing on mature leaves, effectively distancing neonates from P. putida-enriched tissues, may serve as an adaptive life-history interpretation to mitigate the detrimental effects of the bacterium on the F1 generation while still ensuring necessary maternal colonization. Furthermore, the observation that P. putida increases larval mortality without affecting adult survival, fecundity, or egg viability (Figs. 4I and S6) reveals a pronounced ontogenetic shift in immune competence or physiological tolerance. This stage-dependent susceptibility suggests that adults may function as stable, resilient vectors for the bacterium, whereas larvae remain vulnerable. Such ontogenetic variation mirrors broader patterns observed in other insect–pathogen systems. For instance, developmental transcriptomics in Spodoptera frugiperda has uncovered dynamic immune remodeling across life stages41, while studies in Tribolium castaneum confirm that life stage strongly influences both immune gene expression and infection resistance42. More broadly, immune priming studies suggest that insect immune responsiveness is shaped by the interplay between developmental context and prior microbial exposure, contributing to stage-dependent variation in host defence43. Whether this shift is driven by a more robust adult immune system or distinct physiochemical constraints within the adult gut—which may suppress the bacterium’s pathogenic potential—remains to be elucidated. Additionally, a previous study demonstrated that egg-surface P. putida can degrade chlorogenic acid (CGA)44, suggesting a potential protective role for neonates. How these diverse, stage-specific functions of P. putida—ranging from potentially pathogenic in new leaves to protective on egg surfaces—are spatio-temporally synchronized with the host’s life-history stages warrants further investigation. Overall, this tripartite interaction between P. versicolora, P. putida, and the willow host suggests a potential trade-off in which maternal behavior appears to balance the requirements of gut bacteria transmission with the necessity of offspring survival. To pinpoint how P. putida influences oviposition preference, we performed a comparative transcriptome analysis of the heads of GF versus CONV females before oviposition. Fortunately, this analysis revealed a relatively small set of only 61 DEGs. Interestingly, most immune-related DEGs (AMPs) are aberrantly upregulated in GF females compared to CONV controls (Supplementary Data 2). In insects, the gut microbiota plays a key role in maintaining immune homeostasis by modulating basal immune signaling pathways45. Evidence from diverse systems suggests that the depletion of commensal cues can destabilize the homeostatic set point of innate immunity, potentially impairing regulatory mechanisms and leading to dysregulated immune activation46,47. We therefore hypothesize that a similar microbiota-mediated immune calibration occurs in P. versicolora. While this systemic immune activation in GF individuals coincides with the loss of mature-leaf oviposition preference, the precise regulatory mechanisms linking immune status to behavioral output warrant further investigation. Focusing on genes related to neural signaling, we identified a Takeout gene (TO1) that was significantly upregulated in CONV females compared to GF controls. As a member of the Takeout/JHBP family—a group known for integrating nutritional status, circadian rhythms, and endocrine signaling—TO1 aligns with the typical profile of TO proteins, being approximately 250 amino acids long, containing a signal peptide, and possessing a conserved JHBP domain48–50. Given that TO proteins often function as circulating carriers for lipophilic factors like juvenile hormone, TO1 likely serves as a physiological bridge between the host’s internal state and its external behavior. Coupled with TO1 correlated positively with the degree of mature-leaf preference (Fig. 5C), we consequently investigated the function of TO1 in this study. As expected, the knockdown of TO1 disrupted the normal preference for mature leaves (Fig. 5G). Takeout genes were first discovered in Drosophila, where they regulate feeding behavior and circadian rhythms49,50. Subsequent research has frequently linked TO with insect reproduction and physiological adaptation. Specifically, TO is essential for regulating fecundity in Zeugodacus cucurbitae fruit flies51. More intriguingly, TO has been found to be associated with temporal kin recognition in burying beetles52, representing a form of post-recognition selection behavior (though more complex than oviposition preference). Although the underlying mechanism of TO-mediated recognition remains to be fully elucidated, these beetles present a promising model for dissecting how TO influences insect cognitive behavior. To determine if P. putida’s effect on oviposition preference is mediated by the TO1 gene, we conducted experiments to assess whether mono-colonization with cultured P. putida was sufficient to restore the preference phenotype in GF females. We found that P. putida mono-association effectively rescued the mature-leaf oviposition preference in GF females, while the control bacterium did not (Fig. 6B). This behavioral restoration was accompanied by a significant upregulation of TO1 expression specifically in P. putida-colonized individuals (Fig. 6C). Intriguingly, however, P. putida reintroduction remained capable of rescuing the mature-leaf preference even when TO1 was targeted by RNAi (Fig. 6D), suggesting that the behavioral response is robust to transcript reduction and potentially triggered by a low functional threshold of TO1 expression (Fig. 6D). Specifically, the capacity of P. putida to significantly elevate TO1 levels relative to the dsTO1 control—effectively partially overriding the RNAi-mediated suppression—was sufficient to drive the behavioral transition (Fig. 6E). Such partial decoupling between the magnitude of TO1 expression and the behavioral output suggests that TO1 may operate within a threshold-dependent framework. The ability of P. putida to override TO1 knockdown further points to the existence of parallel regulatory pathways—potentially involving other bacterial metabolites or neuromodulatory genes—that can compensate for reduced TO1 availability. For instance, evidence from other insect systems indicates that bacterial metabolites, such as methionine, can modulate host gene expression through RNA m6A methylation53, Whether P. putida utilizes similar epigenetic or metabolic cues to fine-tune the TO1 axis or activate complementary pathways remains a primary objective for future investigation. Furthermore, as TO1 is expressed in both the head and body, our systemic RNAi approach targets both compartments globally; thus, the specific physiological consequences of TO1 knockdown outside the neuronal tissues also warrant further investigation. In addition, we identified 16 additional TO-like genes in the genome (Supplementary Data 2). Although these genes did not show differential expression between the GF and CONV groups, their potential functions remain to be explored. In summary, our study unveils a sophisticated gut bacteria-dependent mechanism shaping oviposition preference in P. versicolora, identifying P. putida as a key microbial partner. We further demonstrate the mechanism by which this bacterium can utilize plant hormone signals (IAA) induced by herbivory to migrate and facilitate its transmission through the host plant. This interaction results in differential impacts on different life stages of the beetle. Although we identified TO1 gene as a potential downstream player responsive to P. putida, the precise molecular signaling pathway linking the bacterium in the gut to this neural response and the resulting behavioral modification requires further elucidation. This work highlights the intricate and dynamic interplay between gut microbiota, plant, and insect behavior in complex ecological systems. Methods Insects, plants, and germ-free treatment Plagiodera versicolora beetles were collected from Sha Lake Park in Wuhan (Wuhan, Hubei, China), Qinhuangdao (Qinhuangdao, Hebei, China), Guiyang (Guiyang, Guizhou, China), and reared in laboratory conditions at 27 ± 1 °C, 70 ± 5% relative humidity, and a 16:8 h light:dark photoperiod. Beetles from each location were fed willow leaves collected from the same site to minimize potential effects of diet switching on experimental outcomes. F1 generation females were used for experiments assessing field oviposition preference, while laboratory-based experiments did not extend beyond the F2 generation. New leaves were defined as the first three leaves below the apical meristem of the plant. Mature leaves were defined as the fourth, fifth, and sixth leaves from the apex16. Conventionally-reared (CONV) P. versicolora adults were used as the control group. Germ-free (GF) adults were obtained by surface sterilization of eggs. Freshly laid egg clutches were collected and placed on humid filter paper. Eggs were surface sterilized with 75% ethanol for 8 min, followed by rinsing three times with sterile distilled water to obtain axenic larvae54. These larvae were then reared on sterile tissue-cultured Salix babylonica plants until adulthood. In addition, we established an antibiotics-treated population (ABX). For this group, leaves were soaked in a mixture of four antibiotics—erythromycin, gentamicin, kanamycin, and tetracycline—each at a concentration of 0.25 mg/mL. Beetles were fed with these treated leaves for two days before being used in subsequent experiments. The efficiency of microbial elimination was verified using the following method, briefly, for culture-dependent verification, intestines were dissected from GF, ABX, and CONV adults under a stereomicroscope (Mshot, Guangzhou, China) within a sterile laminar flow hood. Dissected intestines were homogenized in 300 μL sterile water. A 100 μL aliquot of the suspension was plated onto LB agar medium and incubated at 27 °C for 48 h. For culture-independent verification, homogenized intestines were used for total RNA extraction. Bacterial load was quantified by RT-qPCR using primers specific for the 16S rRNA gene (Supplementary Data 1). Oviposition site choice and bacterial rescue experiments Field oviposition preference experiments were conducted on willow trees at Sha Lake (Wuhan, Hubei, China) that showed no visible insect damage and were mature individuals. To avoid potential interaction between test sites, selected trees were spaced at least 30 m apart. On each tree, five non-adjacent branches were randomly chosen. A pair of gravid P. versicolora adults, freshly collected from the field, was placed on each selected branch, and their movement was restricted using a fine mesh sleeve cage (Fig. 1B). After 24 h, oviposition sites were recorded. The same procedure was applied to antibiotic-treated adults. This experiment was performed for beetle populations from Qinhuangdao, Wuhan, and Guiyang. Throughout the experiments, nearly 100% of the 35 pairs per group successfully laid eggs, from which 30 samples were randomly selected for statistical analysis (n = 30). Given that antibiotic treatment appeared to have minimal impact on oviposition and mating rates, we further evaluated whether it influenced the feeding behavior of ABX females, which could indirectly affect their oviposition preference. For each of the three geographical locations (Wuhan, Qinhuangdao, and Guiyang), four females were randomly selected from the CONV and ABX groups. Each beetle was placed in a Petri dish and provided with fresh new leaves. After 24 h of feeding, the leaves were photographed, and the consumed leaf area was quantified using ImageJ software (v.1.51j8) (n = 4). To further confirm the field results and investigate the underlying mechanisms, the Wuhan population was used to generate germ-free (GF) females for laboratory oviposition assays. In these assays, each dish contained one mature leaf and two to three new leaves, with their combined area approximately matching that of the mature leaf. Leaves were placed on humidified filter paper at the dish base to maintain humidity. A single pair of sexually mature adults (5 days post-eclosion, either CONV or GF) was introduced per dish. After 24 h, oviposition site choice was assessed by counting the number of eggs on each leaf type (mature vs. new). Notably, as nearly 100% of both GF and CONV females successfully initiated oviposition within the 24-h period, the oviposition rate was not quantified as a separate variable. Twenty replicates, each containing six females, were analyzed (n = 20). Preference Index (PI) was calculated for oviposition using the formula: PI = (A − B)/(A + B), where A represents the total number of beetles on mature leaves, and B represents the corresponding total measurement on new leaves26. The mature leaf (ML) oviposition rate was determined by dividing the number of eggs deposited on mature leaves by the total egg count, calculated according to the formula: ML oviposition rate = (number of eggs on mature leaves)/(number of eggs on new leaves + number of eggs on mature leaves). Females that did not lay egg clutches or laid clutches outside the leaves were excluded from the analysis. Egg clutch size was recorded for each female that laid eggs on either new or mature leaves (n = 30 clutches analyzed for size). Egg hatch rate was recorded after 3 days of incubation for each clutch (n = 30 clutches analyzed for hatch rate). To investigate the role of gut bacteria in adult oviposition preference, rescue experiments were conducted. Gut contents were obtained by dissecting and grinding the intestines of two CONV female adults. Pseudomonas putida (accession number: MT791338) and Enterobacter aerogenes (accession number: MT835160) used for rescue experiments were isolated from the gut contents of CONV P. versicolora adults40. Gut contents or the isolated bacteria were cultured overnight at 37 °C. Bacterial suspensions were then resuspended and diluted to 106 CFU/mL in sterile water for subsequent rescue experiments. S. babylonica leaves were coated with 104 CFU/cm2 of bacterial suspension. GF female adults were fed these coated leaves for 2 days before conducting oviposition site choice experiments as described above (n = 20). Egg clutch size and hatch rate were also recorded (n = 30). Gut and leaf microbiota analysis Newly eclosed female adults were fed on either new or mature S. babylonica leaves. Following oviposition preference experiments (n = 20, each containing six females), female adults were surface sterilized with 75% ethanol for 15 s, rinsed three times with sterile water, and their intestines were dissected under sterile conditions. For each treatment (new leaf feeding, NLF, n = 7; mature leaf feeding, MLF, n = 5), each replicate consisted of six pooled intestines from adult females. Leaf samples were collected from wild willow trees in the Jiangxia (new leaf, NL_JX; mature leaf, ML_JX), Wuchang (new leaf, NL_WC; mature leaf, ML_WC), and Huangpi (new leaf, NL_HP; mature leaf, ML_HP) Districts of Wuhan. At each site, five healthy, mature trees (showing no visible insect or disease damage) spaced at least 10 m apart were randomly selected (n = 5). From each tree, three branches were randomly chosen, and one new leaf and one mature leaf were harvested from each branch. The three new leaves and three mature leaves from each tree were pooled separately into single samples and transported to the laboratory on ice for DNA extraction. Prior to DNA extraction, leaf surfaces were gently rinsed three times with sterile water to remove loosely attached environmental contaminants. Bacterial DNA was extracted from intestinal and leaf samples using a FastDNA Spin Kit for Soil (MP Biomedicals, USA) according to the manufacturer’s protocol. DNA quality was assessed by electrophoresis on a 1% agarose gel and visualized under UV light. DNA concentration was determined using a NanoDrop ND-2000 spectrophotometer (NanoDrop, USA). The beetles and leaves 16S rRNA gene were amplified using primers 338F and 806R and 799F/ 1193R, respectively (Supplementary Data 1), and PCR products were purified using a Gel Extraction Kit (Cwbio, China). The resulting amplicon libraries were sequenced on an Illumina Nextseq2000 platform by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Raw reads were quality-filtered using fastp (v0.19.6) and merged using FLASH (v1.2.7)55,56. Subsequently, the high-quality sequences were de-noised using the DADA2 plugin within the Quantitative Insights Into Microbial Ecology version 2 (QIIME2, version 2024) pipeline with recommended parameters57. Taxonomic assignment of amplicon sequence variants (ASVs) was performed using the Naive Bayes consensus taxonomy classifier implemented in QIIME2 and the SILVA 16S rRNA database (v138). Microbial diversity (alpha diversity metrics) and community composition (Principal Coordinates Analysis based on Bray-Curtis dissimilarity, PCoA) were analyzed at the ASV level using the vegan package in R (4.3.2). A Permutational Multivariate Analysis of Variance (PERMANOVA) was used to test for significant differences in microbial community structure between treatments, based on the weighted UniFrac distance matrix. To visualize the taxonomic composition, the top 15 most abundant bacterial genera were selected to generate relative abundance bar plots at the genus level. Furthermore, to characterize the shared and unique microbial taxa among plant tissues and insect guts, bacterial genera from new leaves and mature leaves across the three geographical locations in Wuhan were pooled. These datasets were then compared with the gut microbiota of females fed on new leaves or mature leaves using UpSet plots. To identify key bacterial genera, correlation analysis was performed in R between oviposition data (proportion of eggs laid on mature leaves) and the relative abundance of bacterial genera (n = 12). For absolute quantification of the identified bacterial genera, a separate set of pooled gut samples was collected using the same procedure as described above (n = 9 per group, total n = 18 for the correlation). Absolute abundance was quantified using standard curves generated from known copy numbers of target DNA. Briefly, PCR products amplified with genus-specific primers (Supplementary Data 1) were ligated into the pMD19-T vector (Takara, Japan) and transformed into competent DH5α E. coli cells. Positive colonies were selected and cultured overnight in LB liquid medium supplemented with 50 µg/mL ampicillin. Plasmid DNA was extracted using a SuperPlasmid Mini Kit (Cwbio, China), and concentration was measured using a Nanodrop ND-2000 spectrophotometer (Nanodrop, USA). Plasmid copy numbers were calculated from the DNA concentration, and serial tenfold dilutions were prepared to establish standard curves for qPCR. Correlation analysis, as described above, was also performed between oviposition data and the absolute abundance of specific bacterial genera quantified by qPCR. Selective isolation and identification of Pseudomonas from the female gut To isolate Pseudomonas species, guts from sexually mature CONV females were dissected, and homogenized in 1 mL of sterile PBS (n = 5). The resulting homogenate was serially diluted and spread onto Pseudomonas CFC (Cephalothin-Fucidin-Cetrimide) Selective Agar, a specialized medium for the selective recovery of Pseudomonas species39. After incubation at 28 °C for 48 h, the total number of colony-forming units (CFUs) was recorded. Representative single colonies exhibiting characteristic Pseudomonas morphology were selected and subcultured in LB broth at 28 °C with shaking at 200 rpm overnight. Genomic DNA was extracted using the Bacteria Genomic DNA Extraction Kit (Tiangen, Beijing, China) according to the manufacturer’s instructions. The 16S rRNA gene was amplified using the universal primers 27F and 1492R (Supplementary Data 1). The PCR reaction was performed under the following conditions: initial denaturation at 95 °C for 2 min; 35 cycles of 98 °C for 10 s, 58 °C for 30 s, and 72 °C for 20 s; followed by a final extension at 72 °C for 5 min. The resulting amplicons were purified and sequenced by a commercial service (Sangon Biotech, Shanghai, China). Sequences were compared against the NCBI GenBank database using the BLASTn algorithm to determine taxonomic identity. Based on the abundance and representative nature of the isolates, P. putida was selected for all subsequent re-association and functional assays. Pure cultures were stored in 25% glycerol at −80 °C for long-term preservation. IAA, JA, and CGA content measurement and bacterial chemotaxis assay To investigate the spatiotemporal dynamics of plant indole-3-acetic acid (IAA), jasmonic acid (JA), and chlorogenic acid (CGA) induced by P. versicolora feeding, S. babylonica plants were subjected to treatments in which a single female adult—either conventional (CONV), P. putida-supplemented (CONV + P. p), or antibiotic-treated (ABX)—was confined to feed on the third leaf below the apical meristem (Fig. 3B). Additional treatments included mechanical wounding (MW) at the same leaf position with topical application of ~40 μL (104 CFU/mL) of P. putida (MW-P. p) (n = 4–5). To enhance the abundance of gut P. putida in CONV + P. p group, female beetles were initially fed willow leaves coated with 104 CFU/cm² P. putida for 2 days to obtain P. putida-supplemented beetles. Mechanical wounding was standardized using a hole puncher to create an injury area of approximately 3.0 cm2, matching the average 24-h feeding capacity of a single female (Fig. S1E). Following 24 h of confinement to the third leaf from the apex (Beetle-Damaged Leaf, BDL, Fig. 3B), insects were removed and samples from the BDL, Upper Leaves (UL) were collected at 6, 12, or 48 h post-removal, while MW timing commenced immediately following injury. Tissues were homogenized in cold PBS (0.1 g/mL, n = 4–5), and concentrations were quantified using specialized ELISA kits for IAA (Jining Bio, Shanghai, China), JA (Yuanju Bio, Shanghai, China), and CGA (Jingmei Bio, Jiangsu, China). Furthermore, the abundance of P. putida in these laboratory-collected leaf tissues was quantified via qPCR using P. putida-specific primers (Supplementary Data 1). These findings were validated through field sampling in Wuhan, Qinhuangdao, and Guiyang, where UL from infested plants and healthy controls were analyzed for IAA content (n = 6, one leaf per replicate) as well as the abundance of Pseudomonas genus (n = 6, one leaf per replicate) and P. putida species via qPCR (n = 5, one leaf per replicate) (Supplementary Data 1). P. putida or E. aerogenes cultures were grown at 37 °C with shaking until OD600 reached 0.4–0.5. Cells were harvested by centrifugation at 1536 × g for 5 min and resuspended in PBS buffer. A 200 μL aliquot of the cell suspension was placed in each well of a 96-well plate. A 2 μL capillary tube (HIRSCHMANN®, Germany) was heat-sealed at one end and filled with either PBS buffer (control, IAA, methyl jasmonate [MeJA], or CGA) solution at various concentrations30. The open end of the capillary tube was carefully submerged in the bacterial suspension. After 30 min incubation, the capillary tube was rinsed with sterile water and transferred to a buffer. The bacterial suspension from within the capillary tube was serially diluted and plated on LB agar medium. Plates were incubated overnight at 37 °C, and colony forming units (CFU) were counted. Chemotaxis was quantified by comparing the number of cells entering the IAA, MeJA, or CGA -filled capillary to the number of cells entering the control capillary (filled with PBS). Five replicates were performed for each IAA, MeJA, or CGA concentration. Plant-mediated horizontal and transgenerational transmission of P. putida To facilitate tracking, an engineered P. putida strain expressing enhanced green fluorescent protein (eGFP) was constructed (hereafter referred to as P. putida-eGFP) by amplifying the eGFP gene and inserting it into the broad-host-range plasmid pBBR401 via homologous recombination, which was then introduced into P. putida cells by electroporation and selected on LB agar plates supplemented with gentamicin at 50 μg/mL58. To determine whether P. putida transmits through plant tissues, we performed fluorescence localization using the P. putida-eGFP strain. Following 24 h of feeding by colonized beetles and an additional 12 h of plant cultivation, cross-sections of the nearest upper leaves and longitudinal sections of the internodes between the beetle-damaged leaves and the upper leaves were collected and fixed in FAA solution for 24 h. Fixed tissues were dehydrated through a series of ethanol (75, 85, and 90 for 2 h each; followed by two changes of 100% ethanol for 45 min each), cleared in an ethanol/xylene mixture (1:1) for 10 min and twice in 100% xylene for 8 min each, and finally infiltrated with paraffin (three changes for 1 h each). Tissues were embedded in paraffin at −20 °C, and 4 μm sections were cut using a Leica RM2016 microtome (Leica Microsystems, Wetzlar, Germany). Sections were flattened in 40 °C water, mounted on glass slides, dried at 60 °C. Prior to staining, sections were deparaffinized in xylene and rehydrated through a decreasing ethanol series. After DAPI staining, slides were scanned using a Pannoramic MIDI scanner (3DHISTECH, Budapest, Hungary). Fluorescence signals were detected using standard filter settings for eGFP and DAPI, with acquisition parameters consistently adjusted to minimize plant autofluorescence. To evaluate horizontal transmission, UL tissues collected from plants previously fed upon by P. putida-eGFP-colonized females were provided as the sole food source to newly eclosed female adults for 3 days. Beetle guts were subsequently dissected for RNA extraction (n = 4, 6 guts per replicate), and the presence of the eGFP strain in both plant tissues and guts was quantified via qPCR using eGFP-specific primers (Supplementary Data 1). Furthermore, to assess transgenerational transmission, newly eclosed beetles were allowed to feed on the colonized Upper Leaves (UL) derived from plants previously infested by P. putida-eGFP females until oviposition. Following egg-laying, eggs were transferred to fresh, healthy willow leaves, and the resulting offspring were reared through to adulthood. DNA was subsequently extracted (n = 4) from their F1 offspring at various developmental stages, including eggs (100 per replicate), 1st (6), 2nd (3), and 3rd (2) instar larvae, pupae (1), and adult mouthparts (8), guts (3), and feces collected over a 24-h period from five adults (n = 4). Absolute quantification of the eGFP gene was performed via qPCR to track the persistence and transmission of the bacterial load across generations. Impact of P. putida on larvae performance, pathogenicity, and feeding behavior Larval fitness was assessed by feeding first-instar larvae on UL or LL leaves collected from plants previously fed upon by CONV or ABX adults, or from unfed control plants (Unfed plants). Larvae were reared individually on these leaves until pupation. Body weight of 3rd instar larvae (n = 30) and pupae (n = 30), and developmental time from 1st instar to pupation (n = 30) were measured. Concurrent with these assays, P. putida abundance within UL and LL tissues was quantified via 16S rRNA qPCR (n = 6, one leaf per replicate, Supplementary Data 1), and four larvae were randomly selected at the pre-pupal stage for total RNA extraction and bacterial quantification (n = 4). To further evaluate the virulence of P. putida across different life stages, survival rates were monitored daily for six days (n = 30) in newly emerged larvae (P.p.-larvae) and newly eclosed female adults (P.p.-adult) continuously fed on S. babylonica leaves coated with a 104 CFU/cm2 P. putida suspension, with CONV-larvae and CONV adults serving as controls (n = 30). For the adult groups, a single male was included to ensure mating and was replaced upon death; female fecundity and egg hatching rates were also recorded daily. To determine whether larval P. versicolora can detect the presence of P. putida and consequently exhibit feeding preferences, feeding preference was investigated using a choice assay in 60 mm diameter Petri dishes. Each dish contained four discs of 104 CFU/cm2 P. putida-coated leaves and four control discs (6 mm diameter) placed equidistantly on humidified filter paper. Starved second-instar larvae were placed in the center, and feeding choices were recoded after 12 h (n = 15, each containing four larvae). Preference Index (PI) was calculated using the formula: PI = (A − B)/(A + B), where A represents the number of larvae recorded on P. putida-coated discs, and B represents the corresponding number of larvae on control leaves. RNA extraction and RT-qPCR Total RNA was extracted from tissue samples using Vezol reagent (Vazyme Biotech Co., Ltd, China) according to the manufacturer’s instructions. RNA concentration and purity were quantified using a Nanodrop ND-2000 spectrophotometer (Nanodrop, USA). Complementary DNA (cDNA) was synthesized from 2 μg of total RNA using a Strand cDNA Synthesis for qPCR Kit (Vazyme Biotech Co., Ltd, China) with random primers. RT-qPCR was performed on a CFX96 Real-Time System (Bio-Rad, USA). Each 10 μL reaction mixture contained 5 μL SuperStar Universal SYBR Master Mix (Cwbio, China), 2 μL cDNA template, 0.4 μL of each primer (10 ng/μL), and 2.2 μL sterile double-distilled water (ddH₂O). The PCR conditions were as follows: initial denaturation at 95 °C for 3 min, followed by 40 cycles of 95 °C for 10 s and 60 °C for 30 s. Cq values were obtained after the reaction, and relative gene expression levels were calculated using the 2−ΔΔCt method. The P. versicolora 18S rRNA gene and S. babylonica actin gene (Sa-actin) were used as internal reference genes for insect and plant samples, respectively. Primer sequences are listed in Supplementary Data 1. Transcriptome analysis Head tissues were collected from CONV and GF female adults prior to oviposition by separating the head from the rest of the body below the thorax. The chitinous cuticle was removed from the heads prior to sample preparation (n = 3, with each replicate consisting of pooled head tissues from 10 females). Total RNA was extracted as described above. From 3 µg of total RNA per replicate, cDNA libraries were constructed using standard protocols and sequenced on an Illumina NovaSeq X Plus platform with a paired-end 150 bp strategy. The raw paired-end reads were trimmed and quality-controlled using fastp56 with default parameters. Clean reads were aligned to the reference genome (GCA_037013635.1) in a strand-specific mode with HISAT2 software26,59. The mapped reads of each sample were assembled and quantified using StringTie in a reference-based approach60. Principal Component Analysis (PCA) was performed on the log2 -transformed TPM values of the top 2000 most variable genes across all samples to evaluate group-specific variations. All calculations and visualization were executed in R (v.4.3.2). Differential expression analysis between CONV and GF groups was conducted using DESeq261; genes with |log₂ fold change| ≥ 1 and FDR < 0.05 were considered significantly differentially expressed. Volcano plots were generated to visualize these differentially expressed genes (|log2FC| ≥ 1, FDR < 0.05, TPM > 10). Phylogenetic and spatio-temporal expression analysis of candidate genes Initial transcriptomic profiling identified two DEGs encoding putative takeout proteins (TO1 and TO2). Subsequent analysis revealed that these two sequences share exceptionally high sequence identity ( ~ 99.3%). To determine whether they represent assembly-derived redundancies, locus-specific read depth was assessed using raw genomic data (BioProject: PRJNA1045849). Clean genomic short reads (second-generation sequencing) were re-mapped to the reference genome using BWA-MEM (v.0.7.19-r1273), and the per-base sequencing depth was extracted using SAMtools (v.1.22.1) depth. To establish a 1× genomic coverage baseline, we utilized the median depth of 2,785 “Complete” single-copy orthologs identified by BUSCO (v.5.4.3) against the insecta_odb10 database. Both TO1 and TO2 (131× and 160×, respectively) exhibited significantly reduced coverage compared to the genomic median coverage of 261.7×. To further validate this at the molecular level, we conducted genomic qPCR using a bona fide two-copy gene, Pver038700, as a positive control (Supplementary Data 1). The relative genomic dosages of TO1 and TO2 were 0.73- and 0.76-fold, respectively, compared with Pver038700. Collectively, these bioinformatic and molecular lines of evidence confirmed that TO2 was a misassembled redundancy; it was therefore streamlined into a single gene model (TO1) and excluded from all subsequent functional analyses. Following the identification of the takeout gene, its protein sequence was used for phylogenetic analysis. Representative takeout sequences from key insect orders, including Coleoptera [Phaedon cochleariae (AFQ22726.1) and Tribolium castaneum (XP_974890.2)], Lepidoptera [Helicoverpa armigera (XP_049698879.2)], Orthoptera [Schistocerca nitens (XP_049799292.1)], Diptera [Drosophila melanogaster (NP_001287525.1)], Hymenoptera [Apis cerana (AWC67396.1)], and Hemiptera [Aphis gossypii (XP_050060523.1)] were obtained from NCBI. Phylogenetic trees were constructed using the neighbor-joining method in MEGA7 with 1000 bootstrap replicates. For the spatio-temporal expression analysis of two transcriptome-identified genes, conventional (CONV) female adults were collected at different developmental and reproductive stages: Emergence group: within 12 h post-eclosion; Pre-oviposition group: sexually mature females exhibiting visible abdominal swelling prior to oviposition; Post-oviposition group: within 12 h post-oviposition. RT-qPCR was performed on head (n = 3 or 4, with each sample containing 10 heads) and body tissues (n = 3 or 4, with each sample containing 3 body tissue) collected from these groups at the indicated time points to measure the expression levels of takeout1 (TO1) and neuropeptide like (NPL) using their respective primers (Supplementary Data 1). To further investigate the relationship between oviposition preference and gene expression, head RNA was extracted from MLF (n = 9) and NLF (n = 10) females (6 heads per replicate). Pearson’s correlation analysis was performed between ML oviposition rates and the relative expression of TO1, NPL, and identified immune genes (Attacin 1–3, Defensin, Coleoptericin, and PGRP) (Supplementary Data 1, Fig. S7). RNA interference and bacterial supplementation To investigate the effect of TO1 on oviposition preference, TO1 expression was silenced by feeding adults with dsRNA. DNA templates for dsRNA (dsTO1 and dsGFP) synthesis were generated using primers containing T7 promoter sequences (Supplementary Data 1). PCR products were purified and used for dsRNA synthesis using a T7 RNAi Transcription Kit (Vazyme, China) according to the manufacturer’s instructions. The quality and concentration of synthesized dsRNAs were assessed by gel electrophoresis and spectrophotometry (NanoDrop ND-2000, Thermo Scientific, USA). CONV adults were fed detached and trimmed S. babylonica leaves coated with dsRNA. A dsRNA solution (10 ng/μL) was applied evenly to the leaf surface at a rate calculated to achieve a final concentration of 40 ng/cm² (dsGFP as control, dsTO1). Oviposition preference experiments were conducted as described above after feeding adults dsRNA-coated leaves for 2 days (n = 20 for CONV+dsGFP; n = 10 for CONV+dsTO1). Following the oviposition preference experiments, a random subset of 40 adults from the dsGFP group and 40 adults from the dsTO1 group were collected, along with 40 GF females. RNA was extracted from head tissues of these samples for subsequent qPCR analysis to assess the expression levels of TO1 in CONV insects following TO1 knockdown, using GF females as a reference group (n = 4, each containing 10 heads). To assess potential off-target effects of dsTO1, a bioinformatic prediction was conducted using the Blast Zone tool in TBtools (v2.390). The TO1 gene sequence served as the query against a local database built from the P. versicolora genome-predicted transcripts. Following the established screening criteria, any transcript with a continuous alignment length ≥ 19 bp was identified as a high-risk off-target candidate. To ensure the reliability of the validation, candidates that were not detectable by qPCR in the control (CONV) samples (Ct values > 35 or no amplification) were excluded from further analysis. Based on this analysis, all potential takeout homologs were identified, and their expression levels were quantified via qPCR to ensure the specificity of the RNAi treatment (Supplementary Data 1 and Fig. S8). To evaluate the influence of P. putida supplementation on TO1 expression levels and oviposition preference, bacterial re-introduction experiments were conducted using GF adult females. Additionally, we performed experiments combining bacterial re-introduction with RNAi-mediated TO1 knockdown in GF adult females. For bacterial re-introduction into GF adults, axenic Salix babylonica leaves were coated with suspensions of P. putida or E. aerogenes (used as a control bacterium) at a concentration of 104 CFU/cm². A negative control treatment consisted of leaves coated with an equivalent volume of ddH₂O. Beetles were fed these treated leaves for 2 days. Following the feeding period, beetles were used in oviposition preference experiments as described above (n = 10 or 20, each containing 6 females). After the oviposition assay, a random subset of 36–60 adults from each treatment group was collected for RNA extraction, cDNA synthesis, and subsequent qPCR analysis of TO1 expression (n = 6 or 10, each pooling 6 heads). For experiments combining re-introduction with RNAi, GF adult females were fed leaves simultaneously coated with dsRNA (dsGFP or dsTO1 at 40 ng/cm²) and bacterial suspensions (P. putida or E. aerogenes at 104 CFU/cm²). An additional control group received leaves treated with dsTO1 combined with an equivalent volume of ddH₂O. Beetles were fed these leaves for 2 days before being used in oviposition preference experiments (n = 10, each containing 6 females). After the oviposition assay, a random subset of 24–48 adults from each treatment group was collected for TO1 expression analysis by qPCR (n = 4–8, each pooling 6 heads). To further investigate whether increasing the bacterial load could reinforce existing behavioral patterns, supplementation experiments were also performed using conventional (CONV) beetles. Upon reaching sexual maturity, newly eclosed CONV female adults were supplemented with P. putida (CONV + P.p) or ddH2O(CONV + H2O) via coated leaves as described above. Subsequent oviposition preference tests (n = 60), gut microbial quantification (n = 4, 3 guts per replicate), and head TO1 expression analysis (n = 4, 8 heads per replicate) were conducted to evaluate the reinforcement of the neuroendocrine and behavioral responses. Data analysis Field oviposition preferences were analyzed using the two-sided chi-square test. Differences in adult oviposition preference index in Petri dishes and P. putida attraction or repellency assays were analyzed using a two-sided one-sample t-test (comparing the mean to zero). For comparisons of means between two groups, two-sided independent samples t-tests were performed. For comparisons among three or more groups, one-way ANOVA was conducted, followed by Tukey’s post hoc multiple comparisons. For survival experiments, the significance of differences between survival curves was assessed using the Log-rank (Mantel–Cox) test. Detailed statistical methods specific to the transcriptome analysis (e.g., DESeq2 and PCA) and microbial community analysis (e.g., PERMANOVA and Pearson’s correlation) are described in their respective sections. Differences were considered statistically significant at p < 0.05. Statistical analyses were performed using SPSS version 26 and R version 4.3.2. Graphs were prepared using GraphPad Prism 8, R version 4.3.2 with the ggplot2 package, and Adobe Illustrator 2020. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Acknowledgements We thank Jianyang Bai (Anhui Agricultural University), Qian Zhao (Fujian Agriculture and Forestry University), and Weilong Kong (Huazhong Agricultural University) for their technical assistance with data analysis. We are also grateful to Fangyuan Zhou (Qilu University of Technology) for his insightful suggestions and constructive feedback on the conceptual framework of this study. Author contributions L.X. designed research; C.L., Z.S., Y.Z. and M.M. performed experiments; C.L., Z.S., Y.Z. and L.X. analyzed data. C.L., Z.S. and L.X. wrote and revised the paper. Peer review Peer review information Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This research was supported by National Natural Science Foundation of China (32370523). Data availability The sequencing data generated in this study have been deposited in the NCBI BioProject database under accession code PRJNA1288721 and PRJNA1288147. Source data underlying Figs. 1–6 and all Supplementary Figs. have been deposited in the Figshare database under DOI: 10.6084/m9.figshare.31841368. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Chong Li, Ziyue Sun. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-73654-4. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Data Availability Statement The sequencing data generated in this study have been deposited in the NCBI BioProject database under accession code PRJNA1288721 and PRJNA1288147. Source data underlying Figs. 1–6 and all Supplementary Figs. have been deposited in the Figshare database under DOI: 10.6084/m9.figshare.31841368.